Join The Tinkerer on Whop. Membership gets you early releases, private tools and a bunch of extra stuff.
👉 Join on Whop
💬 Join the community for support, free tools and early news on Discord
CyberHub Lite is now also available on GitHub: https://github.com/cyberdeliaAI/CyberHub-Lite
CyberHub is a local-first command center for AI image creators. The Lite edition is free and works on its own. Point it at your output folders and you've got a complete tool to browse, search and make sense of every image you've ever generated. No account, no cloud, no uploads. It runs entirely on your own machine over localhost, so your images and prompts never leave your PC.
This isn't a stripped-down demo. Lite is a real, self-contained app that does its job completely. The Full edition adds production tools on top, but nothing here is crippled or time-limited.
What's inside
Gallery — Browse huge collections fast. You get a folder tree, favorites, manual collections, auto-generated thumbnails and a clean metadata panel for every image. It stays smooth on galleries of 100k+ images.
Powerful search — Full-text search over prompts, filenames, models and metadata, backed by a SQLite index so it stays instant even on enormous libraries. It also does fragment search on checkpoint names. Type zit and it finds CyberRealistic_zit_v5.0, even though the fragment sits in the middle of the name.
Reads almost every metadata format — Automatic1111 / Forge, ComfyUI, Civitai, InvokeAI (all variants), NovelAI, SwarmUI, Fooocus / RuinedFooocus / FooocusMRE and EasyDiffusion. Positive prompt, negative prompt, sampler, steps, CFG, seed, model, LoRAs and VAE all land in a tidy panel. There's a Raw view too when you want the unfiltered blob.
Built-in Civitai model lookup — Ships with a local models.json database, so many model hashes resolve straight to readable checkpoint names. You can keep it current from Settings whenever you like.
Viewer & Compare — Full-screen viewing plus side-by-side comparison of 2 to 4 images. They don't have to be next to each other. Ctrl+Click any images from anywhere in your gallery to line them up.
Prompt Library — Save your best prompts as cards, attach the source image and build a personal anthology of what actually works. You can send any image's metadata straight from the Gallery into a Library card in one click.
Private by design
Everything runs locally over http://localhost. No telemetry, no accounts, no cloud. Optional LAN mode lets you open the hub from another device on your own network, gated behind an access token. Works on Windows, macOS and Linux.
Install in a minute
Unzip
cyberhub-lite.zipRun
start.bat(Windows) or./start.sh(macOS/Linux). It sets up Python and downloads fonts automatically on first run.Open the URL it prints, go to Settings, add your image folder. Done.
Requires Python 3.10–3.13. Your image files are never moved or modified. The hub treats your folders as read-only sources.
Want the full studio?
Lite gives you everything you need to organize and search. The Full edition turns CyberHub into a complete production studio and adds the tools that do the heavy lifting:
Civitai Grabber — bulk-download images by username or model version ID into clean, flat folders
Danbooru Toolkit + Auto-Tagger — tag lookup, prompt cleanup and local ONNX image auto-tagging, fully offline
AI Captioner — caption your images with a local Vision Language Model
Prompt Engineer — generate, rewrite and convert prompts with LLM assistance and ready-made presets
Cropper, Overlay & Amateur Photo Studio — smart aspect-ratio cropping, watermarks and photo-realistic filters
Meta tools — copy and transfer metadata between images
The Full edition is for members. Same codebase, every module unlocked, free updates.
👉 Get CyberHub Full on Whop: https://whop.com/cyberdelia-ai-lab
Description
Initial release. Free Lite edition: Gallery, Viewer, Compare and Prompt Library.
FAQ
Comments (23)
ur a fkn legend. this looks great cant wait to try it out!
How much the full version costs? Is it a only-one payment or it's a monthly subscription?
It's a monthly subscription ($7/month) through The Tinkerer membership, not a one-time payment. On top of CyberHub Full you get early access to new model releases, private tools and members-only content. Any build you've downloaded keeps working after you cancel, you just stop getting updates. Lite is free.
Based on my experience, I highly recommend a membership.
yeah buddy this is what we're talking about
Full version, LMS runs in background as server on port 1234, but Prompt Engineer says "Detected Model: Connection failed". Developer log in LMS looks good, JSON response visible. (?)
Try switching the server address between:
http://localhost:1234
http://127.0.0.1:1234
Make sure you do not add /v1 at the end, because Prompt Engineer adds that automatically.
This usually fixes it when CyberHub and LM Studio are running on the same machine.
If you are opening CyberHub from another computer, use the IP address of the machine running LM Studio instead, for example:
If it still fails, open the browser console with F12, click Detect Model again, and send me the exact error. If the error mentions CORS, enable the CORS option in LM Studio's server settings and allow the CyberHub address/port.
For a future update, I will likely route Prompt Engineer through the CyberHub server, like Captioner already does. That should remove this browser CORS issue completely.
Got it, inside LMS "enable CORS" (Server settings).
The full version is not this one, is it? Do I need to purchase a membership in this version to get the full version?
No, this is CyberHub Lite, which is the free version. It's fully functional on its own for browsing, searching and reading metadata of your images.
CyberHub Full is separate. It comes with The Tinkerer membership and adds the production tools (Captioner, auto-tagger, Civitai Grabber, Prompt Engineer, and more). So yes, to get Full you'd join the membership.
@Cyberdelia Is the validity period determined by the duration of the membership purchased? So, when the membership expires, the software is no longer used, right?
@Cyberdelia Another question is: What content can be unlocked by becoming a member? Do you provide the download of all the models? Are there any other permission unlocks available?
@Mr_Fei No, the software keeps working even after your membership ends. CyberHub runs entirely on your own machine, so any build you've downloaded stays fully usable. What you lose when the membership ends is access to new updates and new builds, not the version you already have.
You get early access to new releases (early access), plus some members-only models that aren't posted for free. On top of that you get the tools, member only guides and CyberHub Full.
What license is CyberHub Lite released under? I didn't see a LICENSE file or any license header in the source, so I'm not sure what's permitted beyond personal use. Would you mind adding a license to the next release? It'd be really helpful for people who want to contribute, package it, or build around it - I'm currently containerizing it in Docker for personal use / testing.
Great question, and thanks for the nudge!
CyberHub Lite is licensed under the PolyForm Noncommercial License 1.0.0, and a license file will be included (tomorrow hopefully)
In plain terms: you're free to use, modify, share, and containerize it (your Docker setup for personal use/testing is totally fine), and contributions are very welcome. The only thing it doesn't allow is commercial use/resale - so personal, hobby, research and non-commercial projects are all good to go.
Appreciate you wanting to build around it - let me know how the Docker packaging goes!
@Cyberdelia Thanks. I have it working in the Docker container. It's pretty good, metadata extraction works better than CivitAI & Tensor.art. Really nice work :)
I am looking at adding in an API connection to send generation data to Forge for txt2img (maybe others after). a GitHub repo would be useful so I could send PRs for anything added
I actually have this working now for outputs from Forge. Under the 'Save to Prompt Library' I added a 'Generate with Forge' button which pushes the metadata via API to a configured endpoint. The hub then saves the received image to a configured directory and displays it. Round-trips produce the same image. ComfyUI metadata parsing is a bit trickier, needs some love
@psychologicau Thanks so much, really glad it's working well in Docker!
"Generate with Forge" button is awesome, exactly the kind of direction I'd love to see CyberHub go.
I'm setting up a public GitHub repo for CyberHub Lite so you can send PRs. I'll drop the link here once it's up.
Would love to look at your Forge integration as a PR. Funny thing, both the Forge and ComfyUI hooks were already on my list. Which by now has grown into a seriously long list. So many ideas, so little time. 😅
Love it! ;)
When I tried to run ./start.sh in linux, I got an error message about python not being installed, but that is nonsense because I have all the python. I opened start.sh in a text editor and deduced the following install routine:
Manual linux installation of Cyberhub
This is what worked for me
1 decompress the archive.
2 Move the resulting directory to wherever you want to run Cyberhub from.
3 Within that directory, open a command prompt.
4 type: python3 -m venv venv
5 type: source venv/bin/activate
6 type: pip install -r requirements.txt
7 if pip suggests updating itself, then type: pip install --upgrade pip
8 to run cyberhub, type: python3 hub.ph
9 open your browser to 127.0.0.1:8899
10 to close the server/program, go back to the command prompt and press CTRL C
Will include this information in the next/update version. I will also try to update start.sh (if possible). Thanks for sharing the information.
Thank you for making this tool. I was using SD Prompt Reader to see which parts of random prompts were used in images, but Cyberhub is much faster thanks to the gallery view.